> Define weirdness as the Kolmogorov complexity of the shape, within the precision needed to work properly.
Well, if the shape is the output of an optimization program, as long as the program itself and its inputs can be simply specified, the shape actually has low Kolmogorov complexity.
True, but you can still formalize the sense in which the object is weird in terms of how long it takes the optimization routine to run. See Bennett's notion of Logical Depth: https://en.wikipedia.org/wiki/Logical_depth
Well, if the shape is the output of an optimization program, as long as the program itself and its inputs can be simply specified, the shape actually has low Kolmogorov complexity.